Kalshi had a rough 48 hours, and the interesting part is that neither half of it was about odds.
On 10 August, FlightAware filed suit against Kalshi in the Southern District of New York, case 1:26-cv-06824, alleging trademark infringement, breach of contract, unfair competition and reputational harm. The core complaint is not that Kalshi mispriced anything. It is that in mid-July, Kalshi filed a self-certification with the CFTC naming FlightAware as the primary source agency for settling contracts on the share of scheduled flights cancelled at given airports over given periods. FlightAware says nobody asked it whether it wanted to be the referee.
Then on 11 August, with New York suing for at least $36 billion and seeking a restraining order to push Kalshi’s event contracts out of the state, the CFTC invoked Section 8a(9) of the Commodity Exchange Act and ordered Kalshi to keep operating. That is emergency authority, used for only the second time since 1980.
TL;DR
- FlightAware sued Kalshi on 10 August 2026 over flight-cancellation markets, after Kalshi self-certified FlightAware to the CFTC as the primary source agency that settles those contracts.
- On 11 August the CFTC invoked emergency powers under Section 8a(9) to keep Kalshi trading in New York, the second use of that authority since 1980.
- Neither fight is about whether the market is honest. Both are about who owns the answer, and who decides whether the venue stays open.
- The source that settles you is the fifteenth unasked half of fairness. A Chainlink VRF proof certifies that an outcome was generated without interference. It says nothing about outcomes that are observed and reported by a third party.
- Satoshie’s raffles and coinflips have no source agency at all. Randomness is the only input, the proof is verified on-chain by the Base contract, and there is no external party to sue, licence or lobby.
Give the model its due first
Prediction markets are good at what they claim to do. Putting money behind a forecast is a better truth serum than a poll, and Kalshi has spent years grinding through a regulatory process most of crypto refuses to touch. It is a CFTC-registered exchange, it publishes its rules, and it told the regulator exactly where the settlement number comes from. That is more disclosure than a typical crypto casino has ever offered about anything.
The failure here is not dishonesty. It is architectural, and it is the same failure that shows up every time a game’s outcome depends on something that happens outside the game.
The fifteenth unasked half: the source that settles you
Fifteen instalments into this series and the pattern has not changed. Provable fairness is a narrow claim wearing a wide word. Chainlink VRF answers exactly one question, properly: was this outcome generated without interference by anyone, including the operator. The coordinator verifies the proof on-chain before the callback executes, so no node can grind for a preferred number no matter how much money is sitting on it.
But that guarantee only applies to outcomes that are generated. A prediction market outcome is not generated. It is observed. Somebody has to look at the world, count the cancelled flights, and publish a number that closes the position. That step is not cryptography. It is a person, or a company, or a data product with a licence agreement and a commercial roadmap and a legal department.
Kalshi’s phrasing is perfect and nobody noticed it: primary source agency. The market designated an outside company as the entity whose word ends the bet. The number that decides who gets paid is a private product, owned by a firm that now says the arrangement breached its terms and misused its trademarks. Every trader in that market was trusting FlightAware. Almost none of them thought of it that way, because the trust was buried in a self-certification filing rather than presented as what it is: a dependency.
This is the unasked half. Ask a platform whether its draw is fair and you might get a real answer. Ask it who settles the bet, what happens if that source changes methodology, revokes access, reclassifies a delay as a cancellation, or simply refuses to participate, and you will usually get silence. Outcome integrity and resolution integrity are different properties, and only the first one is provable.
A referee that never agreed to have skin in the game
FlightAware’s second argument is sharper than the trademark claim, and it deserves attention beyond aviation. The company argues that markets on cancellation rates create a financial incentive for someone to influence cancellations.
Think about what that means structurally. The moment real money settles on a real-world measurement, the measurement stops being a neutral observation and becomes a target. The source did not opt into that. FlightAware built a flight tracker, not a settlement layer for an event-contract exchange, and it has now been handed the liability profile of a referee without any of the protections a referee would negotiate.
Every crypto game that settles on an external feed has this shape. Price oracle, sports API, weather data, in-game leaderboard. The randomness might be beyond reproach while the input that actually decides the payout sits with a party who has their own incentives, their own outages, their own lawyers, and no contractual duty to you at all.
The venue is a political variable too
The CFTC order is the third instalment of this series repeating itself. Liveness is not correctness. Kalshi’s markets stayed open on 11 August not because the code kept running but because a federal agency decided a state lawsuit was a market emergency. Good for its users this week, and a reminder that availability was always a decision made in a room they were not in.
Why Satoshie has no source agency
Here is the design choice, stated plainly. A Satoshie raffle and a Satoshie coinflip have exactly one input, and that input is randomness. There is no external number to fetch. No API to licence. No agency to certify to a regulator. Nothing that a third party can withdraw, reprice, redefine or sue over.
The randomness comes from Chainlink VRF, and the proof is verified by the contract on Base before the result exists. Everything that decides who wins happens inside the smart contract, on a chain anyone can read, using an input whose honesty is a mathematical property rather than a commercial relationship. If Satoshie’s front end vanished tomorrow, the draw logic and its history would still be there and still be checkable.
That is not cleverness. It is a deliberately smaller claim, and the smallness is the feature. Fewer inputs mean fewer parties who can be wrong, absent, compromised or unwilling.
The honest limit
Satoshie cannot tell you how many flights got cancelled at Newark. It will never run a market on an election, a football result, or the weather, because the moment a game depends on an observed fact it inherits every problem this article describes, and no amount of verifiable randomness patches that.
So the takeaway is not that prediction markets are broken. It is that if you are playing a game whose outcome depends on data somebody else owns, “provably fair” cannot reach the part that matters most. Ask who settles the bet. Ask what happens when they decline. If the answer is a company that found out about the arrangement in a court filing, you were never playing a fair game. You were playing a fair draw, attached to somebody else’s business decision.
📷 Photo by Joseph Bobadilla on Unsplash
